How much does site speed affect conversion rate?
Measurably, and most steeply in the first few seconds - but the exact figure is store-specific. Bucket your own sessions by load time and compare conversion rates instead of quoting industry stats.
Public research consistently shows slower pages convert worse, with the sharpest drop-off in the first few seconds of load time and on mobile. What no public study can give you is your slope: audience, device mix, price point, and traffic source all move it.
The defensible method: segment sessions by load-time bucket, compare conversion rate and average order value across buckets, and value the sessions you could realistically move into faster buckets. That yields a dollar figure your CFO can audit, unlike a borrowed stat.